Latest news with #Gemini1.5Flash
Entrepreneur
14-05-2025
- Business
- Entrepreneur
Google Cloud Ramps Up AI Infrastructure in India, Eyes Local Model Deployment
The company is actively partnering with the Indian government on the INR 10,000 crore IndiaAI Mission to provide subsidised compute power to startups and researchers You're reading Entrepreneur India, an international franchise of Entrepreneur Media. Google Cloud is deepening its artificial intelligence (AI) investment in India, a market it considers critical to its global growth strategy. The company plans to eventually bring the full spectrum of its AI stack — including advanced models, proprietary hardware, and development platforms into the country. While the Gemini 1.5 Flash model is currently hosted in India, more advanced versions like Gemini 2.5 Flash are expected to follow soon as part of Google's 12-month AI refresh cycle. Local model hosting will offer Indian enterprises and developers reduced latency, enhanced performance, and greater data sovereignty. "India as a market, the growth that we are seeing is one of the fastest in the world and hence, we are moving very, very quickly to see how do we not just augment capacity but augment technology," said Bikram Singh Bedi, Vice President and Country Managing Director, Google Cloud India. "So that the customers in India can then use the latest parts of our technology." Google Cloud currently operates out of two cloud regions in Mumbai and Delhi NCR. The Delhi data centre region was developed with a strategic focus on supporting government and regulated industries. The company is actively partnering with the Indian government on the INR 10,000 crore IndiaAI Mission to provide subsidised compute power to startups and researchers. Among its public sector collaborations, Google Cloud is powering the iGOT Karmayogi platform, which uses AI to offer personalised learning pathways for civil servants. The company's end-to-end AI stack includes its proprietary Vertex AI platform, TPUs, GPUs, and a suite of out-of-the-box agents designed for tasks such as customer service, security, and data analysis.
Business Mayor
10-05-2025
- Science
- Business Mayor
Fine-tuning vs. in-context learning: New research guides better LLM customization for real-world tasks
Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More Two popular approaches for customizing large language models (LLMs) for downstream tasks are fine-tuning and in-context learning (ICL). In a recent study, researchers at Google DeepMind and Stanford University explored the generalization capabilities of these two methods. They find that ICL has greater generalization ability (though it comes at a higher computation cost during inference). They also propose a novel approach to get the best of both worlds. The findings can help developers make crucial decisions when building LLM applications for their bespoke enterprise data. Fine-tuning involves taking a pre-trained LLM and further training it on a smaller, specialized dataset. This adjusts the model's internal parameters to teach it new knowledge or skills. In-context learning (ICL), on the other hand, doesn't change the model's underlying parameters. Instead, it guides the LLM by providing examples of the desired task directly within the input prompt. The model then uses these examples to figure out how to handle a new, similar query. The researchers set out to rigorously compare how well models generalize to new tasks using these two methods. They constructed 'controlled synthetic datasets of factual knowledge' with complex, self-consistent structures, like imaginary family trees or hierarchies of fictional concepts. To ensure they were testing the model's ability to learn new information, they replaced all nouns, adjectives, and verbs with nonsense terms, avoiding any overlap with the data the LLMs might have encountered during pre-training. The models were then tested on various generalization challenges. For instance, one test involved simple reversals. If a model was trained that 'femp are more dangerous than glon,' could it correctly infer that 'glon are less dangerous than femp'? Another test focused on simple syllogisms, a form of logical deduction. If told 'All glon are yomp' and 'All troff are glon,' could the model deduce that 'All troff are yomp'? They also used a more complex 'semantic structure benchmark' with a richer hierarchy of these made-up facts to test more nuanced understanding. 'Our results are focused primarily on settings about how models generalize to deductions and reversals from fine-tuning on novel knowledge structures, with clear implications for situations when fine-tuning is used to adapt a model to company-specific and proprietary information,' Andrew Lampinen, Research Scientist at Google DeepMind and lead author of the paper, told VentureBeat. To evaluate performance, the researchers fine-tuned Gemini 1.5 Flash on these datasets. For ICL, they fed the entire training dataset (or large subsets) as context to an instruction-tuned model before posing the test questions. The results consistently showed that, in data-matched settings, ICL led to better generalization than standard fine-tuning. Models using ICL were generally better at tasks like reversing relationships or making logical deductions from the provided context. Pre-trained models, without fine-tuning or ICL, performed poorly, indicating the novelty of the test data. 'One of the main trade-offs to consider is that, whilst ICL doesn't require fine-tuning (which saves the training costs), it is generally more computationally expensive with each use, since it requires providing additional context to the model,' Lampinen said. 'On the other hand, ICL tends to generalize better for the datasets and models that we evaluated.' Building on the observation that ICL excels at flexible generalization, the researchers proposed a new method to enhance fine-tuning: adding in-context inferences to fine-tuning data. The core idea is to use the LLM's own ICL capabilities to generate more diverse and richly inferred examples, and then add these augmented examples to the dataset used for fine-tuning. They explored two main data augmentation strategies: A local strategy: This approach focuses on individual pieces of information. The LLM is prompted to rephrase single sentences from the training data or draw direct inferences from them, such as generating reversals. A global strategy: The LLM is given the full training dataset as context, then prompted to generate inferences by linking a particular document or fact with the rest of the provided information, leading to a longer reasoning trace of relevant inferences. When the models were fine-tuned on these augmented datasets, the gains were significant. This augmented fine-tuning significantly improved generalization, outperforming not only standard fine-tuning but also plain ICL. 'For example, if one of the company documents says 'XYZ is an internal tool for analyzing data,' our results suggest that ICL and augmented finetuning will be more effective at enabling the model to answer related questions like 'What internal tools for data analysis exist?'' Lampinen said. This approach offers a compelling path forward for enterprises. By investing in creating these ICL-augmented datasets, developers can build fine-tuned models that exhibit stronger generalization capabilities. This can lead to more robust and reliable LLM applications that perform better on diverse, real-world inputs without incurring the continuous inference-time costs associated with large in-context prompts. 'Augmented fine-tuning will generally make the model fine-tuning process more expensive, because it requires an additional step of ICL to augment the data, followed by fine-tuning,' Lampinen said. 'Whether that additional cost is merited by the improved generalization will depend on the specific use case. However, it is computationally cheaper than applying ICL every time the model is used, when amortized over many uses of the model.' While Lampinen noted that further research is needed to see how the components they studied interact in different settings, he added that their findings indicate that developers may want to consider exploring augmented fine-tuning in cases where they see inadequate performance from fine-tuning alone. 'Ultimately, we hope this work will contribute to the science of understanding learning and generalization in foundation models, and the practicalities of adapting them to downstream tasks,' Lampinen said.
Time of India
10-05-2025
- Business
- Time of India
Google Cloud to continue ramping up AI push in India
Google Cloud will continue to increase its capacity in India, which it considers a critical market, as it expects to bring in all layers of its artificial intelligence (AI) stack - across models and hardware - within the country over time, a top executive said in New Delhi on Friday. When it comes to models, so far, only the Gemini 1.5 Flash model is hosted in India, but with 12-month refresh cycles, newer versions will also be made available locally, said Bikram Singh Bedi , vice president and country MD at Google Cloud India. 'India as a market, the growth that we are seeing is one of the fastest in the world and hence, we are moving very, very quickly to see how do we not just augment capacity but augment technology, to bring in the latest and the greatest so that the customers in India can then use the latest parts of our technology,' Bedi said. Currently, Google Cloud has infrastructure capacity in two data centre zones in India - Mumbai and Delhi NCR. In a significant move last October, Gemini 1. 5 Flash became the first Google model enabled to run locally here to enable greater control and security for customers. Since then, however, more advanced versions such as Gemini 2.5 Flash - said to be more efficient as well - have been released. On the ₹10,000 crore IndiaAI Mission , under which the government is providing subsidised AI compute to startups and researchers, Bedi said that the company is working closely with the government on the opportunity and is keen to be a part of it. Google Cloud's Delhi region was set up with an eye on serving the public sector, he added. The company has been deepening its engagements with government agencies. For instance, it is partnering to host the civil servant capacity-building platform iGOT Karmayogi where it enables personalised recommendations, among other use cases. Bedi said Google can optimise price performance of AI solutions better because all four layers of the stack are owned by it - the infrastructure, research and models, AI agents, and its AI training and deployment platform Vertex AI. The company has released out-of-the-box AI customer agents, security agents, creative agents, and data agents, which can carry out tasks for these functions autonomously. It is also building AI agents that can interoperate with third-party AI agents. The Google Cloud platform offers AI graphics processing units (GPU) and tensor processing unit (TPU) chips, and Google's Gemini, Imagen, Chirp, and Lyria multimodal models, among others.
Time of India
09-05-2025
- Business
- Time of India
Google Cloud to continue ramping up AI push in India
Google Cloud will continue to increase its capacity in India, which it considers a critical market, as it expects to bring in all layers of its artificial intelligence (AI) stack - across models and hardware - within the country over time, a top executive said in New Delhi on Friday. #Operation Sindoor India-Pakistan Clash Live Updates| Missiles, shelling, and attacks — here's all that's happening Pakistani Air Force jet shot down in Pathankot by Indian Air Defence: Sources India on high alert: What's shut, who's on leave, and state-wise emergency measures When it comes to models, so far, only the Gemini 1.5 Flash model is hosted in India, but with 12-month refresh cycles, newer versions will also be made available locally, said Bikram Singh Bedi , vice president and country MD at Google Cloud India. 'India as a market, the growth that we are seeing is one of the fastest in the world and hence, we are moving very, very quickly to see how do we not just augment capacity but augment technology, to bring in the latest and the greatest so that the customers in India can then use the latest parts of our technology,' Bedi said. Currently, Google Cloud has infrastructure capacity in two data centre zones in India - Mumbai and Delhi NCR. In a significant move last October, Gemini 1. 5 Flash became the first Google model enabled to run locally here to enable greater control and security for customers. Since then, however, more advanced versions such as Gemini 2.5 Flash - said to be more efficient as well - have been released. On the Rs 10,000 crore IndiaAI Mission , under which the government is providing subsidised AI compute to startups and researchers, Bedi said that the company is working closely with the government on the opportunity and is keen to be a part of it. Discover the stories of your interest Blockchain 5 Stories Cyber-safety 7 Stories Fintech 9 Stories E-comm 9 Stories ML 8 Stories Edtech 6 Stories Google Cloud's Delhi region was set up with an eye on serving the public sector, he added. The company has been deepening its engagements with government agencies. For instance, it is partnering to host the civil servant capacity-building platform iGOT Karmayogi where it enables personalised recommendations, among other use cases. Bedi said Google can optimise price performance of AI solutions better because all four layers of the stack are owned by it - the infrastructure, research and models, AI agents, and its AI training and deployment platform Vertex AI. The company has released out-of-the-box AI customer agents, security agents, creative agents, and data agents, which can carry out tasks for these functions autonomously. It is also building AI agents that can interoperate with third-party AI agents. The Google Cloud platform offers AI graphics processing units (GPU) and tensor processing unit (TPU) chips, and Google's Gemini, Imagen, Chirp, and Lyria multimodal models, among others.
